Why Do Educated Mothers Matter? A Model of Parental Help
Transcription
Why Do Educated Mothers Matter? A Model of Parental Help
Discussion Paper No. 2010-29 | December 17, 2010 | http://www.economics-ejournal.org/economics/discussionpapers/2010-29 Why Do Educated Mothers Matter? A Model of Parental Help Luciano Canova Enrico Mattei School Alessandro Vaglio University of Bergamo Abstract This paper investigates the role performed by mothers in affecting their childrens’ performance at school. The article develops firstly a theoretical model in which household (parent–child pair) is treated as an individual, whose utility depends both on the performance at school of the student and on consumption. The model focuses on the different possibilities through which help of mothers may affect pupil’s performance both in terms of time devoted to supervision and spillover effects. Empirical evidence then, using PISA 2006 and focusing on Italian case, shows that education of mothers is an issue when interacted with her occupational status. Highly educated mothers have a positive impact on students’ score only when they are highly qualified in the job market. Paper submitted to the special issue Quasi Markets in Education JEL J12, J24, I21 Keywords Education; PISA; quantile regressions; parental help Correspondence Luciano Canova, Via S.Salvo 1, 20097 S.Donato Milanese, Italy; e-mail: [email protected] © Author(s) 2010. Licensed under a Creative Commons License - Attribution-NonCommercial 2.0 Germany 1 Introduction Achievement is a multidimensional concept. An individual may remain poor but reach a respected position in the community he lives in, or he may earn a high income in an ill-famed profession, have a good school or army record, or again he simply may succeed in living up to his/her standards, whatever they are. In this paper we discuss a particularly important type of achievement, namely school achievement , and the relationship between children’s achievement and their parents’education level As Haveman and Wolfe (1995) put it: "(...) perhaps the most fundamental factor [among those in‡uencing child’s educational attainment] is the human capital of parents, typically measured by the number of school years attained. This variable (...) is included in virtually every study described [in their paper] (...);it is statistically signi…cant and quantitatively important, no matter how it is de…ned." In this paper we do not challenge this conclusion, but try to look in more depth into it. The question is, as the title of the paper says, why is parents’education so important in children performance. We begin by concisely reviewing two issues related to the in‡uence of the family on children performance, while the third one lies at the core of the paper. The …rst issue is the well known "nature or nurture ?" question. Parents pass on to their children their genetic endowment. To the extent that school achievement depends on genetic factors, there might be no genuine causal relationship running from parents’to children’s school achievements, both being explained by the common genetic endowment. The idea that "genetics explains all", has far-reaching implications for the very conception of equality among human beings and for public policies pursuing such equality.This explains why this issue has been the subject of recurring harsh public debates (see for instance Jensen (1969), Kamin (1974), Herrnstein-Murray (1994)) . The typical methods of behavioural genetics (i.e. studies on twins and adoptees) have been used not only by genetists but also by scholars with a social and economic orientation (see Bound-Solon (1999), Holmlund, et al. (2008) and Pronzato (2008) for comparison among methods). As a whole, it seems that although genetic factors are non-negligible, they account for a partial share of the individual’s performance (see for instance Thompson et al.(1991) and DeFries et al. (2001)). Once the relative weights of nature and nurture are established, much remains to be said about what "nurture" is and how it works. The next important question regards the relative weights of family- versus non-family in‡uences. Even in disciplines usually very keen on within-family e¤ects, like developmental psychology, the issue of non-family e¤ects emerges (see for instance the heathed debate stimulated in the psychologists’…eld by Harris(1998)). On the 2 other hand, even if one is willing to stress the in‡uences coming from the social environment at large, neverthless one would agree that such in‡uences are possibly mediated by the family environment (see also Patacchini-Zenou (2004) for a choice-based approach to the issue). Almost all recent studies based on the OECD -PISA database (see in particular Fertig-Schmidt (2002) , Fertig (2003), Bratti et. al. (2007), Sprietsma (2006), Fuchs and Woessman (2004) , Ammermullaer (2004)) include and emphasize background variables like territorial factors or the immigrant condition as explanatory variables. In this paper too a wide array of these background variables is taken into account. The third important question is the nature of within-family in‡uences. As economists, we are particularly interested in distinguishing between in‡uences which require a speci…c allocation of resources and those which do not (see Becker and Tomes (1986) for an early contribution). Day-by-day interaction within the family in‡uences per se the performance of children : just think of how transmission of ethical values, through moral suasion or personal example may induce or not children to put e¤ort into studying; of how critical events such as death or divorce or long-term factors like the number of siblings can a¤ect the school performance of children (Woessman (2003)). Notice also that, in these examples, the in‡uence is in principle independent of the parents’human capital and, always in general, it may not require any speci…c allocation of time or other resources to the task. In other cases it is still true that no time/resource allocation is needed for the in‡uence to take place, but human capital may instead be required. This happens with within-family knowledge spillovers: the simplest example is how children may learn a correct and sophisticated use of language by everyday exchanges with correctly speaking parents. There are, …nally, forms of in‡uence of parents on the school achievement of children which require time . These include on the one hand supervision activities, which per se are independent of parents’human capital (like for instance controlling that children do their homework or regularly meeting teachers); on the other hand there are assistance activities which require both time and human capital (for instance, helping children when studying) . Which of the in‡uences described is more or less relevant to child’s school performance is an issue of some policy import. Were supervision time the sole factor in‡uencing school performance, increasing the level of education of parents would be not bene…cial (on this particular respect) , since it would increase the opportunity cost of time for parents and reduce their propensity to spend time in supervising children. If instead knowledge spillovers were the key factor, the conclusion would be the reverse one. Then understanding why is parents’ education important may help in designing long-term education policies. There is a rather sensitive point to be made here. In this paragraph and in the model of the next one, we speak of "parent" and "parents" without any speci…cation of gender; all we say applies in principle to mothers and fathers alike. The economics of education literature is however rather keen on this kind of distinction . (some recent contributions are Holmlund et al. (2008), Pronzato (2008), Behrman et al. (1997), Andrabi et al. (2009)). There are good reasons for this 3 interest. The main one is that any notion of development meaning more than GDP growth, includes improving women education levels as a critical goal; then the e¤ects of women education vìs-a-vìs male education become crucial. The other one is that that women and men play di¤erent roles within family.The nature and the strength of social and cultural background determines what such roles are and how strict or interchangeable thay are; to the extent that di¤erent roles are systematically played, the same education level in mother and father might have di¤erential impact on the children performance. As we shall see later, our data seem to con…rm that the impact of mother’s education is more complex than that of father’s. However we do not o¤er explanations of this di¤erence. We simply take it as given and discuss at some length the evdience concerning mother’s education. In this paper we model the parents’choice about the allocation of time between labour and children "guidance", which includes both supervision (guidance time whose e¤ectiveness is independent of parents’ education ) and assistance (guidance time whose e¤ectiveness might in principle be sensitive to parents’ education). Parents are characterized by a level of education and by an occupational status, which are observable, and by an underlying unobserved characteristic (parents’quality) which is related to education and to the occupational status in a non-deterministic way. (Section 2). The model provides an interpretative framework for some of the results we obtain in Section 3. Section 4 concludes the paper. 2 A model of parental support Let us consider the household (the parent-child pair) as an individual agent whose utility depends on s (the school achievement of the child,as measured by the test score), and on consumption c. Suppose that the household then maximizes the following utility function max u (c; s) = c1 1 +s c (1) This functional form allows for a decreasing marginal utility of consumption and for some complementarity between consumption and child’s performance. Parents are characterized by two parameters: one, discrete and observable e, the education attainment, which takes values e0 < e1 < e2 . The other, continuous and unobservable, q corresponds to the "quality" of the individual. This is related to education, in the sense that for e = ei (i = 0; 1; 2), the quality q of an individual with ei is q = ei + , where is a random variable with mean 0. Then, if ei > ej : 4 prob fq qb; ei g = 1 F (b q ei ) > 1 F (b q where F (:) is the distribution function of qb; ej g ej ) = prob fq . The parent owns a time endowment T which he/she can employ in the labour market, earning a wage w (q), or in supervising the child in studying. w (q) is a Mincer function with w0 (q) > 0. If we denote as h the time devoted to supervision, the budget constraint is c = w (q) (T h) School achievement s depends on h through the following education production function (Hanushek (2008)) s = kh + b (q) (2) b(q) is a continuous increasing function of q, while k depends on q in the following way kH if q > q0 kL if q q0 k(q) = for some threshold q0 , with kH > kL We therefore assume that the ability of a parent in assisting the children changes with the parent’s quality level, but in a discrete way and only at relatively low levels of education. Then the parent’s quality a¤ects children performance both directly (b (q)) and through a higher e¢ ciency in the use of time devoted to children assistance. Now let us consider an individual with a given quality level q > q0 :After suitable substitutions, the objective function becomes: ' (h; q) = [w (q) (T 1 1 h)] + (kh + b (q)) [w (q) (T h)] (3) where k = kH . The necessary condition for an internal solution is 1 'h (h; q) = nw (q) (T h) + 1 (kh + b (q)) w (q) (T h) k (kh + b (q)) (T Hence: 5 h) 1 o =0 (4) 1 w (q) (kh + b (q)) = (T k (kh + b (q)) h) 1 (T h) (5) 1 For su¢ ciently small changes in q; the full impact on the test score s is given by: ds dh = k(q) + bq dq dq (6) ' hq dh The full expression for dh dq is dq = 'hh , where 'hh < 0; therefore, as q increases (by su¢ ciently small amounts), h changes according to the sign of the following expression + h 1 'hq (h; q) = ( ( 1)k(kh+b(q)) k(kh+b(q)) 1 1 1 + ) w (q) i 1 (T h) 1 w (q) 1 (T h) (T h) (T h) wq w(q) + (kh + b (q)) 1 bq (7) Equation (7)summarizes the two components of the impact of q on h. In the absence of a direct e¤ect (bq = 0), the sign of 'hq depends on ( 1 + ) : If there is no complementarity between consumption an school achievement ( = 0) it is still possible that an increase in e increases the amount of family support (a rapidly declining marginal utility of consumption is required, i.e. > 1): The sign of the second term is also ambiguous. If < 1; it is certainly negative: a higher value of q increases the direct spillovers, so that the parent can reduce the time devoted to children assistance. By means of some tedious calculations, however, it can be shown that the term 1 n 'hh 'hh h ( 1)k(kh+b(q)) k(kh+b(q)) 1 1 (T (T h) h) 1 1 i 1 w (q) (T h) (kh + b (q)) 1 o which represents the overall contribution of the term bq to the variation of s , is always positive. Suppose on the other hand that q changes from some initial level q < q0 to q + q > q0 , where q is …nite. Then we have the following change in the test score h q+ s = kL q q h q q + b q+ q b q q 6 + (kH kL ) h q+ q q (8) While the …rst and the second term in (8) are the discrete counterparts of h(q+ q ) the two terms of (6), the term (kH kL ) , unambiguously positive, q appears only in this discrete case. Now, if we compare two parents,A and B with education levels respectively e0 and e2 ; the probability that qA < q0 and qB > q0 is: F (q0 e0 ) [1 F (q0 e2 )], while if we compare parents C and B , where the education level of C is e1 ; the probability that qC < q0 and qB > q0 is F (q0 e1 ) [1 F (q0 e2 )]. This means that we are more likely to observe the h(q+ q ) when we compare a low level of e¤ect described by the term (kH kL ) q education with higher levels, rather than when comparing levels both relatively high. Unfortunately, we do not observe h nor q, but only e and y, the occupational status of the parent. The occupational status (ybl =blue collar-low quali…ed, ybh =blue collar-high quali…ed,ywl =white collar-low quali…ed, ywh =white collarhigh quali…ed, u = unemployed) is assumed to signal, however imperfectly, both the quality of the parent and the use of time he/she makes.More precisely we assume: prob fe; q qb; ywh g prob fywh g > prob fywh ; eg = prob fywh ; e; q qbg = prob fe; q qbg prob fe; ywh g prob fywh g , i.e. prob feg prob fe; q qb; ywh g prob fe; ywh g > prob fe; q qbg prob feg (9) This assumption simply means that the probability of an individual getting a high quali…ed, white collar job is higher if he/she has a given level of education and he/she is of high quality, than in the case he/she simply has that level of education. Notice that since by de…nition: prob fe; q we have prob fe; q qb; ywh g and prob fe; ywh ; q prob fywh g qb; ywh g = prob fe; q qb; ywh g prob fywh g = prob fe; ywh ; q prob fq qbg We can easily prove the following lemma Lemma 1 Under assumption (9) : prob fq prob fe; q prob fq prob fe; q qb; ywh g prob fq qbg qbg qb; e; ywh g > prob fq (10) qb; eg prob fe; ywh ; q qbg prob fq qbg > prob fq qb; eg = prob fe; ywh g qbg prob fe; ywh ; q qbg prob fe; q qbg that is: > prob fe; ywh g prob feg qb; e; ywh g = qbg prob fq prob feg qbg = 7 Proof. By assumption: we get: that is: prob fe; q qb; ywh g prob fe; ywh g > . Substituting (10) prob fe; q qbg prob feg prob fe; ywh g prob fe; ywh ; q qbg prob fq qbg > prob fe; q qbg prob fywh g prob feg prob fe; ywh ; q bg wh ;q q or: probfe;y > probfe;ywh g qbg > prob fe; q qbg prob feg prob fe; ywh g prob fe; q prob feg qbg According to our assumption, we should then expect that a high level of education associated with a high quali…ed, white collar job is also associated to a high quality level of the parent. Assuming that being unemployed signals quality, i.e. that the probability of being unemployed is larger for low-quality individuals, is certainly too rash; on the contrary, it seems reasonable to assume that being unemployed signals that the opportunity cost of parent’s time is rather low, and therefore it should, ceteris paribus, lead to more assistance to children. 3 Data and method The Programme for International Student Assessment (PISA) is an OECD survey for educational attainment which tests 15 year-old students in the subjects of mathematics, science and reading pro…ciency. We use the 2006 cross-section, which includes data about the 24 OECD countries plus other 33 countries. Along with test scores in reading, math and science, information is collected about many characteristics in a students’and in a school questionnaire1 . The students’questionnaire contains information about family background, socio-economic status in terms of ownership of durable goods, a speci…c focus on science2 issues. The school’s questionnaire contains information about number of students enrolled, number of teachers part-time and full-time employed, quality of infrastructure, type of funds which the school receives (public or private). For a detailed description of PISA dataset, see OECD (2006). Looking at data used for this analysis, the Italian sample consists of 21773 students and 799 schools. Italy is one of the worst achievers among industrialised countries in terms of PISA scores. OECD 2006 is no exception. In this paragraph, the distributions 1 A parents’ questionnaire is available for some of the countries containing information about parents’perception of students’achievement, but we do not use this data 2 It’s the reference theme for 2006 8 of test scores for Italy is presented. Non-parametric kernel density estimates describe the score distribution of the country. Figure 1 displays the test score distributions for reading and math tests in the case of Italy (ITA). In the next few sections we proceed to econometric analysis. The average reading score is 486 points, while for math the average score is 482 points. Figure 1 In order to perform our econometric analysis, we divide the scores into 3 quantiles, de…ned at 0.15, 0.50 and 0.85. The assumption behind this choice is that we can discriminate three levels of students: low performing, average students and high performing students. We then run quantile regressions to catch the e¤ects of a speci…ed set of regressors on the performance of students in di¤erent parts of the distribution, testing di¤erent dependent variables: score in math; overall average score3 ; the overall average score relative to school of attendance. The set of controls does include: age and gender of student; his/her migration status4 ; a variable saying if pupils speak a di¤erent language from that of the 3 Aritmetic mean of math, reading and science tests. consider as migrants the students who were born in a di¤erent country from that of the school they attend 4 We 9 school’s country; school’s type5 , school’s region6 , education of both mother and father7 ; occupational status of both mother and father8 , a composite index for socio-economic status at family level ; the number of books owned at home9 , the lack of quali…ed math teachers; city size10 ; the presence of other school of the same type in the neighbourhood; availability of computers at school; a variable saying if the students studies at home; a variable saying if the students takes out of school lectures. Education is treated as a set of dummy variables saying if the parent has low education, a high school diploma or a degree. The database contains some information about employment status of parents, although not as detailed as we perhaps would like. Since we know if working parents are white or blue collars, high or low quali…ed, we assume that individuals who are not classi…ed neither as blue- nor white collars but that are registered in the dataset are not working. Finally, we introduce an interaction term containing the educational level variable and the indicator of the occupational status. 4 Estimation Results If we look at variables other than parent related ones, we …nd a number of standard result, along with some less conventional suggestions. A general feature is that many coe¢ cients become smaller in absolute value as we consider higher quantiles. The likely meaning of this feature is that adverse factors are compensated for by increasing student’s talent, while favorable in‡uences become less and less crucial to the …nal performance. Unsurprising results are obtained as regards the school’s type, the region and the non-native condition: attending a technical school11 , living in the southern regions and being a nonnative entail lower scores. More precisely, students of lyceums get on average 45 points more than student of technical schools. Students in the north of Italy get on average 30 points more than students in the south. Being a non native student determines a 20-25 lower scores than that of native students. Two less conventional results are: a) the coe¢ cient for spoken language at home is positive, which means that per se speaking a foreign language constitutes an asset (while, as we have seen, being an immigrant is not). 5 Lyceum or technical institute variables says if the school is in the North or in the South of Italy 7 Edum and Eduf in the results 8 Msecateg and Fsecateg in the results 9 The variable is a dummy with di¤erent categories, going from 0 to more than 500 books 1 0 village with less than 3000 people; small town (between 3000 and 15000 people); town (15000 up to 100000 people); city (100000 up to 1milion people); large city (more than 1 milion people) 1 1 Italian high school system is basically split into pre-university institutes, like classical and scienti…c lyceums, and more technical and professional schools 6 The 10 b) Males perform on average better than females in math (which is a commonly found result): however, the coe¢ cient is increasing with the quantile considered, a phenomenon for which an explanation is needed. As regards parents’ education, the coe¢ cients referred to father’s contribution are signi…cant and positive on average. The weight of the coe¢ cient is rather small in comparison to that of the mother, with the expected sign. As regards mothers, three main results are obtained 1. While holding a high school degree has a positive and signi…cant impact on child’s performance, the coe¢ cient associated to graduated mothers is negative and signi…cant. In terms of the model, if we assume that there exists a threshold q0 at which the mother’s skill in assisting children when studying increases to a highel level (kL to kH ), we might interpret this result by saying that the threshold is likely to be reached when the parent earns a high school degree. This explains the strong observed e¤ect. When the parent earns a university degree, kH does not increase anymore. 2. The fact that the mother works has a positive impact on the performance of their child. If anything, this outcome signals that allocation of time to children assistance or supervision is not the key factor for a good child’s performance. 3. When we control for the occupational status of mothers, the interaction between being graduated and holding a high quali…ed job has a positive impact on the performance of pupils, with higher impact for students with a lower score. This is all the more signi…cant if we remind that holding a university degree per se has a negative impact of children’s behaviour. The e¤ect of being a high quali…ed white collar mother translates, on average, into a 20 points higher score of the student. We check the robustness of results by trying di¤erent dependent variables, e.g. the math score and the relative score computed with respect to the school average score. We also ran regressions clustered for school. These estimates con…rm the previously exposed results. We also estimated the same models for di¤erent countries in PISA dataset: namely, Sweden and Portugal. We followed here the work by Ammermueller (2004) and Canova and Vaglio (2010). Sweden is a Scandinavian country which we choose according to traditional development of welfare state of countries within this area, while Portugal, besides very similar to Italy as regards the distribution of the PISA scores, may also be considered similar to Italy with respect of the general structure of the cultural context belonging to the Mediterrenean model of welfare state (Esping-Andersen, 1996) . Portugal yields a set of results very close to those concerning Italy, which comforts us on the robustness of estimates. Swedish data instead produces remarkably di¤erent outcomes. Since the Swedish socio-economic and cultural context is commonly held to be completely di¤erent from the "Mediterranean" 11 model common to Italy and Portugal, it is reasonable to conjecture that these social and economic context variables could be of some help in explaining the peculiar features of mother’s education. Results of these regressions are presented in Appendix A. Table 1 -Quantile regressions: results for 0.15, 0.50, 0.85 quantiles: overall scores VARIABLES sex lyceum employed _Ieduf_2 _Ieduf_3 _Iedum_2 _Iedum_3 _Imsecateg_1 _IeduXmse_2_1 _IeduXmse_3_1 fsecateg migrant_student _Iforeign_l_1 escs _Inr_books_2 _Inr_books_3 _Inr_books_4 _Inr_books_5 _Inr_books_6 no_sch_available _Icity_size_2 _Icity_size_3 _Icity_size_4 _Icity_size_5 school_north study_out selfstudy Observations (1) q15 (2) q50 (3) q85 2.341 (1.904) 59.83*** (1.172) 89.02* (48.24) 11.03*** (1.712) -8.103*** (2.778) 8.843*** (1.594) -25.01*** (6.593) 2.911 (5.144) -3.782 (7.201) 26.21** (10.63) 5.742*** (1.404) -37.67*** (4.172) 20.81*** (1.666) 5.330*** (0.660) 15.45*** (1.845) 30.71*** (2.975) 41.18*** (4.277) 53.08*** (2.393) 60.01*** (6.718) -4.639* (2.558) 15.92*** (5.160) 29.80*** (4.677) 35.01*** (4.785) 27.82*** (9.669) 39.45*** (1.190) -45.58*** (2.988) 10.44*** (2.376) 10.52*** (1.346) 53.06*** (0.792) 70.18*** (15.10) 8.567*** (1.914) -3.238 (2.470) 13.23*** (1.002) -17.99*** (3.281) 5.358** (2.163) -5.562 (3.446) 22.90*** (5.606) 2.817 (1.880) -33.10*** (3.897) 16.05*** (1.978) 3.954** (1.548) 12.98*** (3.244) 30.18*** (1.858) 42.46*** (3.009) 49.30*** (2.815) 61.18*** (4.173) -6.022** (2.971) 21.02*** (6.962) 34.42*** (6.144) 40.29*** (6.330) 36.05*** (6.826) 37.46*** (1.783) -47.23*** (1.267) 4.098* (2.094) 14.23*** (2.315) 42.80*** (0.820) 18.05 (15.37) 12.37*** (1.930) 3.967** (1.727) 9.953*** (2.447) -12.48* (6.950) 4.719 (3.476) -2.724 (3.034) 19.20*** (4.747) 1.304 (3.186) -20.45*** (5.989) 12.21*** (1.175) 0.517 (1.677) 19.92*** (5.128) 38.48*** (3.995) 51.58*** (4.231) 60.40*** (3.343) 70.01*** (1.731) 0.118 (2.645) 18.13** (8.401) 25.38*** (7.698) 33.66*** (5.127) 21.63*** (7.209) 31.86*** (1.800) -50.61*** (3.075) -0.593 (1.225) 15192 15192 15192 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 * denote level of significance: * = 10%; ** = 5%; *** = 1% 5 Conclusions Why is parents’education so important? The general …nding of this paper is that parents’education in‡uences children achievement through indirect channels rather than through relatively time-intensive channels. Motivation, knowledge spillover, provision of a material and emotional environment enhancing school performance seem therefore to be the driving factor. This is particularly 12 important if we think that these results refer to mother’s education and use of time (e.g. women access to the job market), which are two key issues in the discussion over women role in the society: our results mean that an educated and high quali…ed mother bene…ts the school achievement of her children independently, at least to some extent, of the time she devotes to direct supervision and help. Were our conclusions con…rmed by further research, then those institutional arrangements which enhance women education and participation to the job market (and to the most quali…ed segments thereof), would receive further support. The policy relevance of such a conclusion is important in terms of design of social policies aiming at helping mothers through childcare allowances or other public transfer, which, at the same time, do not preclude job perspectives of bene…ciaries. The relevance of the issue is striking for a country like Italy which is traditionally considered an example of the familistic welfare state model (Esping Andersen, 1990) and that experiences an overheated debate about the possibility of reforming radically education system. The main evidence seems to be that quantity of time devoted to education of children is not a driving factor in having an impact on the performance of students. Parents who have invested in human capital have a strong positive impact on students’ scores and the quality of their job is relevant.Spill-over e¤ects or personal example provided by parents is probably playing a role here and must be take into consideration as an evidence that pupils’ educational attainment is the multidimensional result of many factors, requiring therefore targeted public interventions for each of them, References [1] Ammermullaer , A. (2004) "PISA: What Makes the Di¤erence ? Explaining the Gap in PISA Test Scores Between Finland and Germany" ZEW Discussion Papers, n. 4 [2] Andrabi, T. - Das, J. - Khwaja, A.I. (2009) "What Did You Do all Day ? Maternal Education and Child Outcomes" Mimeo [3] Becker,G.S.-Tomes, N. (1986) "Human Capital and the Rise and Fall of Families" Journal of Labour Economics, July, n. 4:S1-39 [4] Bratti,M.-Checchi,D.-Filippin, A. (2007) "Territorial Di¤erences in Italian Students’ Mathematical Competencies: Evidence from PISA 2003", IZA Discussion Papers, n. 2603 [5] Bound, J. -Solon,G. (1999): Double Trouble: On the Value of Twin-Based Estimation of the Return to Schooling. Economics of Education Review, 18, 169-182. 13 [6] DeFries, J.C.- McClearn, G.E-McGu¢ n, P.- Plomin, R. (2001) Behavioural Genetics Worth Publications,New York [7] Esping Andersen G. (1990). The Three Worlds of Welfare Capitalism. Polity Press. Princeton University Press. 1990. [8] Fertig, M. – Schmidt C.M. (2002) "The Role of Background Factors for Reading Literacy: Straight National Scores in the PISA 2000 Study", IZA Discussion Papers, n. 545 [9] Fertig, M. (2003) "Who’s to Blame? The Determinants of German Students’Achievement in the PISA 2000 Study", RWI Discussion Papers, n. 4 [10] Fuchs, .-Woessman, L. (2004) "What Accounts for International Di¤erences in Student Performance?A Re-examination Using PISA Data", IZA Discussion Papers, n. 1287 (NO) [11] Hanushek, E.A (2008) "Education Production Function" The New Palgrave Dictionary of Economics, 2nd Edition, MacMillan. [12] Harris, J.R. (1998) The Nurture Assumption The Free Press, New York [13] Herrnstein, R.J. -Murray C. (1994) The Bell Curve:Intelligence and Class Structure in America, Free Press,NY [14] Holmlund, H., M. Lindahl and E. Plug (2008): Estimating Intergenerational Schooling E¤ect: A Comparison of Methods. IZA Discussion Paper, no 3630 [15] Haveman, R. - Wolfe, B. (1995), “The Determinants of Children Attainments: A Review of Methods and Findings”, Journal of Economic Literature 33:4, pp. 1829-1878. [16] Jensen, A. (1969) "How Much Can we Boost I.Q.and Scholastic Achievement", Harvard Educational Review , February [17] Kamin, L.J. (1974) The Science and Politics of IQ,LEA, Potomac. [18] Patacchini, E - Zenou, Y. (2004) "Intergenerational Education Transmission:Neighborhood Quality and/or Parents’ Involvement?" The Reserach Institute for Industrial Economics, Working Paper No. 631. [19] Chiara Pronzato,C. (2008) Why Educated Mothers don’t make Educated Children? A Statistical Study in the Intergenerational Transmission of Schooling, Statistics Norway, Discussion Papers No. 563, October [20] Sprietsma, M. (2006) "Regional School Comparison and School Choice: How do They Relate to Student Performance" Dèpartement de Sciences Economiques – Louvain Discussion Papers, n. 24 14 [21] Thompson, L.A.-Detterman, D.K.-Plomin, R. (1991) Association between Cognitive Abilities and Scholastic Achievement: Genetic Overlap but Environmental Di¤erences, Psychological Science , 2: 158-165 [22] Wolter S.C. (2003) "Sibling Rivalry: a Six Country Comparison" , IZA Discussion Papers, n. 734 15 Appendix A 16 17 18 19 20 Please note: You are most sincerely encouraged to participate in the open assessment of this discussion paper. You can do so by either recommending the paper or by posting your comments. Please go to: http://www.economics-ejournal.org/economics/discussionpapers/2010-29 The Editor © Author(s) 2010. Licensed under a Creative Commons License - Attribution-NonCommercial 2.0 Germany